Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
ModelZoo
GPT2_pytorch
Commits
2f9cfcec
Commit
2f9cfcec
authored
Jun 16, 2023
by
hepj987
Browse files
调整格式
parent
75e3f49b
Pipeline
#364
canceled with stage
Changes
2
Pipelines
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
4 additions
and
4 deletions
+4
-4
README.md
README.md
+3
-3
run-train.sh
run-train.sh
+1
-1
No files found.
README.md
View file @
2f9cfcec
...
...
@@ -117,12 +117,12 @@ NHEADS 注意力机制头数
SEQ_LEN 最大长度
SAVE_INTERVAL 保存频率
--train
-sample
s 训练
样本
数
--train
_iter
s 训练
步
数
--eval-interval 验证频率
--eval-iters 验证iter
```
### 性能和收敛性
###
16B模型
性能和收敛性
| 卡数 | 性能(samples per second) | 收敛性lm loss value | 收敛性lm loss PPL |
| :-------: | :------------------------: | :-----------------: | :---------------: |
...
...
@@ -197,7 +197,7 @@ sh run-inf.sh(这里以单节点小模型为例)
## loss收敛情况
1
5
B模型使用oscar数据集收敛情况如下:
1
6
B模型使用oscar数据集收敛情况如下:
...
...
run-train.sh
View file @
2f9cfcec
...
...
@@ -15,7 +15,7 @@ CODECARBON_PATH=output_dir/codecarbon/$MODEL_NAME
N_GPUS
=
8
TP_SIZE
=
4
# always fixed to the size of a single node
PP_SIZE
=
2
#128 #96 # NLAYERS must be a multiple of PP_SIZE here
PP_SIZE
=
1
#128 #96 # NLAYERS must be a multiple of PP_SIZE here
MICRO_BATCH_SIZE
=
2
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment